Development of a new Middleware for Real Time Image Processing in Remote Sensing

Abstract

At the German Aerospace Center (DLR) a new project aiming at the efficient control of traffic and at providing help in cases of natural disaster is being developed which is supposed to transmit pre-processed images of high resolution from a plane to a station on the ground. Aboard the aircraft a computer network is installed which is to process these images shot at short intervals. This intensive computational work is done with the latest methods of image processing and is allocated to several modules running on different computers. The process of this computationally intensive work on highly correlated modules requires a permanent exchange of small and big amounts of data of totally different data types. Here a middleware is needed, which guarantees easy and efficient communication between the modules and simultaneously hides their spatial distribution from them. The first part of this thesis analyses to what extent the process of exchange of small and big amounts of data is supported by existing middlewares. Then the new middleware DANAOS, which is developed in this thesis, is presented, a middleware which compared with other middlewares not only guarantees an efficient communication between the modules using message passing but also supports the exchange of big amounts of image data using a distributed shared memory. For this purpose the module programmer is offered an application programming interface, which - in addition to this - provides further services for group communication and name service.